Research map / Label-efficient and robust learning / Recognition and long-tail
Do Explanations Explain? Model Knows Best
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use GradCAM rather than methods with larger null-feature scores | Explanation attribution · Null feature | null-feature metric: 0.835 -> 0.135 | Helped |
| Use Extremal Perturbation rather than methods with larger null-feature scores | Explanation attribution · Null feature | null-feature metric: 1.03 -> 0.047 | Helped |
| Use GradCAM for output-sensitive attribution in the single-feature case | Explanation attribution · Single feature scenario | class-sensitivity metric: 0.979 -> 0.050 | Helped |
| Use GradCAM, IBA, or Extremal Perturbation in the double-feature case | Explanation attribution · Double feature scenario | class-sensitivity metric: 0.931 -> 0.039 | Helped |
| Use Extremal Perturbation for saturated features | Explanation attribution · Feature saturation | feature-saturation metric: -0.680 | Hurt |
| Apply IBA at a later rather than earlier ResNet layer (ablation) | IBA · Null feature | null-feature metric: 0.315 -> 0.211 | Helped |
| Apply GradCAM++ at a later rather than earlier ResNet layer (ablation) | GradCAM++ · Null feature | null-feature metric: 0.827 -> 0.453 | Helped |
| Apply IBA at a later rather than earlier ResNet layer (ablation) | IBA · Double feature scenario | class-sensitivity metric: 0.327 -> 0.191 | Helped |
| Apply GradCAM++ at a later rather than earlier ResNet layer (ablation) | GradCAM++ · Double feature scenario | class-sensitivity metric: 0.977 -> 0.469 | Helped |
Closest papers by meaning
- Evaluating Neuron Explanations: A Unified Framework with Sanity Checks (ICML 2025 · similarity 0.87)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.87)
- Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable? (ICLR 2025 · similarity 0.87)
- Beyond Topological Self-Explainable GNNs: A Formal Explainability Perspective (ICML 2025 · similarity 0.86)
- Factor Graph-based Interpretable Neural Networks (ICLR 2025 · similarity 0.85)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.